73 Using statistical techniques to improve the estimation of quality.

H. M. Allen, P. J. Martin, and A. B. Frensham. NSW Agriculture, Agricultural Institute, Wagga Wagga, 2650 NSW, Australia.

Estimates of early generation quality for breeding lines from trials are obtained using a Quadramat Junior flour mill and other small scale testing equipment. They are known to have errors that are attributable to variation in the field and laboratory. Error variation in the field may follow trends linked to the spatial location of the plots. “Spatial” analysis can accommodate trend in traits such as grain yield that are measured at the field plot level. For a trait such as flour extraction there may also be laboratory trends linked to the order in which samples are milled. Presented will be a method that enables identification and separation of field and laboratory trends. The key aspects are the use of a two-level experimental design, namely a valid field design and a subsequent re-randomisation of samples to ensure that the order of milling differs from field plot order. The statistical analysis is an extension of spatial analysis to accommodate laboratory trend. We present an example in which both field and laboratory trends are evident, the latter dominating. Such trends must be considered when interpreting flour extraction and other quality data. Our method quantifies the trend and removes it statistically, providing adjusted estimates of quality. These estimates was proven to be more accurate and more precise than the unadjusted data for flour extraction.